subnet_bridge: regenerate main.py for explicit local miner loading

main.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import importlib.util
 import json
 import os
 import sys
@@ -8,7 +9,21 @@ from typing import Any
 
 import cv2
 
-from miner import Miner
+
+def _load_local_miner_class():
+    miner_path = Path(__file__).resolve().parent / "miner.py"
+    spec = importlib.util.spec_from_file_location("manako_bridge_local_miner", str(miner_path))
+    if spec is None or spec.loader is None:
+        raise RuntimeError(f"Could not load miner module from {miner_path}")
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    miner_class = getattr(module, "Miner", None)
+    if miner_class is None:
+        raise RuntimeError(f"miner.py does not export Miner in {miner_path}")
+    return miner_class
+
+
+Miner = _load_local_miner_class()
 
 
 CLASS_NAMES = ['football', 'player', 'pitch']
@@ -66,13 +81,42 @@ def load_model(onnx_path: str | None = None, data_dir: str | None = None):
     }
 
 
+def _candidate_keypoint_counts(miner: Any) -> list[int]:
+    counts: list[int] = [0]
+    for attr in ("n_keypoints", "num_keypoints", "keypoint_count", "num_joints"):
+        value = getattr(miner, attr, None)
+        if isinstance(value, int) and value > 0:
+            counts.append(value)
+    counts.append(32)
+
+    seen: set[int] = set()
+    ordered: list[int] = []
+    for count in counts:
+        if count in seen:
+            continue
+        seen.add(count)
+        ordered.append(count)
+    return ordered
+
+
+def _predict_batch_with_fallbacks(miner: Any, image: Any) -> list[Any]:
+    errors: list[str] = []
+    for n_keypoints in _candidate_keypoint_counts(miner):
+        try:
+            return miner.predict_batch([image], offset=0, n_keypoints=n_keypoints)
+        except Exception as exc:
+            errors.append(f"n_keypoints={n_keypoints} -> {exc}")
+            continue
+    raise RuntimeError("predict_batch failed for all keypoint candidates: " + " | ".join(errors))
+
+
 def run_model(model: Any, image: Any = None, onnx_path: str | None = None, data_dir: str | None = None):
     del onnx_path
     if image is None:
         image = model
         model = load_model(data_dir=data_dir)
     miner = model["miner"]
-    results = miner.predict_batch([image], offset=0, n_keypoints=0)
+    results = _predict_batch_with_fallbacks(miner, image)
     if not results:
         return [[]]
     frame_boxes = _extract_boxes(results[0])